LFM2-2.6B vs Llama 3.2 3B Instruct
Wins 1 of 5 areas
Reasoning
Wins 3 of 5 areas
Facts · Long documents · Following instructions
Llama 3.2 3B Instruct is the stronger all-rounder.LFM2-2.6B is better at reasoning.
Scores updated · 21 tests both models report · How we compare
Where each one wins
Tests won in each of the five areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- FactsGetting facts right instead of making them up01Llama 3.2 3B Instruct1 of 2 tests · 1 tie
- Long documentsFinding answers in very long texts01Llama 3.2 3B Instruct1 of 1 test
- Following instructionsDoing exactly what it is asked01Llama 3.2 3B Instruct1 of 1 test
- ReasoningHard problems that need careful thinking10LFM2-2.6B1 of 2 tests · 1 tie
- CodingWriting and fixing software11Even1 each
Long documents and following instructions rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Llama 3.2 3B Instruct pulls ahead
- Follows unfamiliar, precisely checkable instructionsIFBench+6.7points ahead
- Avoids making up answers it doesn't knowAA-Omniscience · Non-hallucination+5.4points ahead
- Code for real scientific research problemsSciCode+2.7points ahead
Where LFM2-2.6B pulls ahead
- Graduate-level biology, physics and chemistry questionsGPQA Diamond+5.1points ahead
- Recent programming contest problemsLiveCodeBench v6+2.9points ahead
Every test, side by side
All 21 tests both models report. The winning score is in its model's colour; marks a score checked independently.
CodingEven
- LiveCodeBench v6LFM2-2.6B by 2.914.411.5+2.9
- SciCodeLlama 3.2 3B Instruct by 2.72.55.2+2.7
ReasoningLFM2-2.6B
- GPQA DiamondLFM2-2.6B by 5.130.625.5+5.1
- Humanity's Last Examtie5.55.3tie
FactsLlama 3.2 3B Instruct
- AA-Omniscience · Non-hallucinationLlama 3.2 3B Instruct by 5.437.142.5+5.4
- AA-Omniscience · Accuracytie5.46.3tie
Long documentsLlama 3.2 3B Instruct
- AA-LCRLlama 3.2 3B Instruct by 4.304.3+4.3
Following instructionsLlama 3.2 3B Instruct
- IFBenchLlama 3.2 3B Instruct by 6.719.526.2+6.7
Other results13 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- HumanEval+LFM2-2.6B by 33.857.924.1+33.8
- MATH-500 (EM)LFM2-2.6B by 22.463.641.2+22.4
- GSMPlusLFM2-2.6B by 22.160.838.7+22.1
- MGSMLFM2-2.6B by 16.174.358.2+16.1
- MMLU-ProLlama 3.2 3B Instruct by 92635+9
- τ²-Bench Telecom (AA run)Llama 3.2 3B Instruct by 7.613.521.1+7.6
- MMMLULFM2-2.6B by 7.555.447.9+7.5
- AA-OmniscienceLlama 3.2 3B Instruct by 6.6-54.1-47.5+6.6
- GSM8KLFM2-2.6B by 4.782.477.7+4.7
- MMLULFM2-2.6B by 464.460.4+4
- LCB v5LFM2-2.6B by 2.914.411.5+2.9
- IFEvalLFM2-2.6B by 2.279.677.4+2.2
- AA Intelligencetie5.25.7tie
Questions people ask
Which is better, LFM2-2.6B or Llama 3.2 3B Instruct?
Llama 3.2 3B Instruct wins three of the five areas where both have results: facts, long documents and following instructions. LFM2-2.6B wins reasoning. They are level on coding.
Which is better for coding?
Neither. They win 1 coding test each of the 2 both models report.
How do you compare the two?
We use the 21 benchmark tests both models have published scores on. The verdict counts the 8 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 13 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.